1 The Model Intuition: Put It M,1M S σ S σ

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

1 The Model Intuition

Our new model will separate the types of ’private’ information that you can capture from option
contracts about the underlying asset. The traditional view (in microstructure model used on
options) is that informed agents can use these derivatives ONLY to exploit their private infor-
mation related to a change (i.e. and its direction) in the fundamental value of the underlying
assets. However, options could be also used to exploit information about the volatility of the
fundamental value.

Because of the structure of option options, informed traders could take profit from their
knowledge on volatility, which CANNOT be used (at least directly) on the underlying equity
markets. This is an important point, since the use of traditional market microestructure models
about information (which has been developed for stocks) could not fit properly over options.
This is because traditional models do not have the volatility information component. In fact,
it is possible that the volatility information component could bias the results of previous stud-
ies about informed trading in option markets (see, e.g., Muravyev (2016, JF)). Moreover, the
inventory risk could be overestimated if volatility is not taken into account in microstructure
models used in option markets.

To explain the intuition of ’our’ new model, suppose we have puts over the same equity,
with different moneyness. For example, suppose we have three option contracts (1-month to
maturity): the put option A which is in-the-money (ITM), the put option B which is at-the-
money (ATM), and the put option C which is out-of-the-money (OTM). Suppose that at time
t we observe a trade in the put option A. This trade has three components: inventory risk
(IN VITP ut S
M,1M ), information about a directional change in the underlying asset (IN F ), infor-
mation about volatility (IN F σ ). It is important to notice that IN F S and IN F σ reflect private
information about the underlying asset, thus they do not depend of the option feautures (i.e.
strike price, time-to-maturity and if contract is a put or a call). Let assume that the depth at the
ask price of the put option A is bigger than the trade size. This example is described in Figure 1.

Figure 1 shows that the ask price of the put option A increases due to the three component:
P ut
IN VIT S and IN F σ . This is because the market maker, who received this order,
M,1M , IN F
learns from this transaction; and thus she updates the ask price. However, it is important to
notice that the effect of the information components, IN F S and IN F σ , are multiplied by the
P ut P ut
absolute value of the option delta (|δIT M,1M |) and the option vega (νIT M,1M ).

Moreover, since the transaction is in the put option A, the ask prices of put options B and C
are not affected by the inventory risk. Nevertheless, because information in modern electronic
markets is standardized, other market makers learn from the trade in the put option A about
the informational components. Thus, other market makers update the ask prices of put options
B and C due to IN F S and IN F σ . However, the effect of IN F S and IN F σ on ask prices of put
options B and C are ’adjusted’ by the respective option deltas and option vegas.

Thus, we can have a linear system with three equations and three unknown variables:

1
∆P utA,ask P ut P ut S P ut
IT M,1M = IN VIT M,1M + |δIT M,1M |IN F + νIT M,1M IN F
σ
(1)
∆P utB,ask P ut S P ut
AT M,1M = |δAT M,1M |IN F + νIT M,1M IN F
σ
(2)
∆P utC,ask P ut S P ut
OT M,1M = |δAT M,1M |IN F + νIT M,1M IN F
σ
(3)

Figure 1: Example of price impact’s components: If a trade occur on put option A, and
the trade has information about directional change in the underlying asset (IN F S ), information
about volatility (IN F σ ), the price impact will reflect the inventory risk (IN VIT P ut
M,1M ) and
S σ
information (IN F and IN F ). The information components will flow to the other put options
o price impacts depending on the delta and vega of the put option contract, the value of these
greeks will depend on the moneyness and time-to-maturities.

Therefore, if we consider different contracts (call and put option contracts), the directions
of the price impact will depend on the delta and vega of each option. Because of the different
value and behaviour of delta and vega depending on the maturity and moneyness, is possible to
extract both types of information, controlling for inventory using trades from different option
contracts.

Our new model is an extension of Muravyev (2016, JF), in which we separate volatility in-
formation and directional information of the underlying asset that is hidden in the price impact
on options contracts.

This separation process also could be made using the characteristics of options markets.
Unlike equity markets, options markets has the characteristics that are several options from
the same equity traded at the same time, separated by their moneyness and maturity. Using
these options that share the same equity is possible to extract the inventory risk using several
options, without requiring been traded in two or more exchanges (as in Muravyev,
2016). This approach could help avoid tick size and exchange characteristics that

2
could bias the results, characteristics that Muravyev (2016) measure has to address.

2 The model
We will the same notation as in Muravyev (2016, JF). In our new model, we have to separate
the inventory risk and the information from each option trade. If at time t, there is a trade on
the put option contract with strike price K and time-to-maturity T , this option contract has a
price impact and will increase the ask price by:

∆P utaK,T,t = E(∆P utaK,T,t | Ft ) + θ|δK,T,t


P ut P ut
|(X − E(xt | Ft )) + φνK,T,t (X − E(xt | Ft ))
+γK,T · X + t+∆t,K,T + ϑt+∆t,K,T − ϑat,K,T
a


Where:

∆P utaK,T,t : is the price impact (percentage) of the ask price of a put with strike K, maturity T
E(∆P utaK,T,t | Ft ): is the expected price impact because of the public information available on time t
P ut
θ|δt,K,T |(X − E(xt | Ft )): is the price impact depending on the information over the underlying
contained in the trade, weighted by the delta of the option
P ut
φνt,K,T (X − E(xt | Ft )): is the price impact depending on the information over the volatility
contained in the trade, weighted by the vega of the option
γK,T · X: price impact depending on the inventory pressure generated by the trade

While non-trading option contracts with strike price K ∗ and time-to-maturity T ∗ get the same
information but have no changes in inventory, then their price impact on the ask price are:

∆P utaK ∗ ,T ∗ ,t = E(∆P utaK ∗ ,T ∗ ,t | Ft ) + θ|δK


P ut P ut
∗ ,T ∗ ,t |(X − E(xt | Ft )) + φνt,K ∗ ,T ∗ (X − E(xt | Ft ))

+t+∆t,K ∗ ,T ∗ + ϑat+∆t,K ∗ ,T ∗ − ϑat,K ∗ ,T ∗




a a Call P ut
∆Callt,K ∗ ,T ∗ = E(∆CallK ∗ ,T ∗ ,t | Ft ) + θδt,K ∗ ,T ∗ (X − E(xt | Ft )) + φνt,K ∗ ,T ∗ (X − E(xt | Ft ))

+t+∆t,K ∗ ,T ∗ + ϑat+∆t,K ∗ ,T ∗ − ϑat,K ∗ ,T ∗




a a Call P ut
∆Callt,K,T = E(∆CallK,T,t | Ft ) + θδt,K,T (X − E(xt | Ft )) + φνt,K,T (X − E(xt | Ft ))
+t+∆t,K,T + ϑt+∆t,K,T − ϑat,K,T
a


Ww can have same equation for bid prices of option contracts. Following Muravyev, to
simplify the notation we will assume that buy and sell trades are conditionally equally likely,
and thus E(xt | Ft ) = 0, and that the noise/error  terms are combined into a single term
a a
t+∆t,K,T = t+∆t,K,T + ϑt+∆t,K ∗ ,T ∗ − ϑt,K ∗ ,T ∗ that has zero mean by construction.

In order to be able to extract the inventory risk and information components we need to
address two elements. First, the microestructure noise t,K,T have to be solved working with
a sufficiently large sample that the microstructure noise terms average to zero. The second

3
element, is the expected changes in price, E(∆P utat | Ft ), which has to be estimated using a
regression (see below).

Notice that (as described above) the big difference between this model and Muravyev’s model
is the separation of the information in two types, volatility and directional change of the under-
lying asset.

3 Expected changes in price


Following Muravyev, we have to generate a regression using public information to predict changes
in prices, in order to decompose the price impact (see the internet appendix on his webpage).
In order to be consistent with this idea, our regression should account for changes on equity and
also on volatility of it, then the regression for the expected quote changes over the ask price,
should be:

12
 X
∆P utaK,T,t = α0 + α1 P utsaK,T,t − µt + P ut

αn+1 |δK,T,t−n∆t | · ∆St−n∆t
1
12
X 12
P us
 X
+ αn+13 νK,T,t−n∆t · ∆σt−n∆t + αn+25 ∆P utaK,T,t−n∆t + t,i
1 1

Where:

∆P P utaK,T,t = ask price impact over an option with price p, on exchange i


µt = mid-price
P ut
δK,T,t−n∆t = Delta of the put option on time t − n∆t
P ut
νK,T,t−n∆t = Vega of the put option on time t − n∆t
σt−n∆t = volatility of the equity on time t − n∆t

3.1 Volatility of equity


There are several ways to measure the volatility of the equity. One choice is to use a GARCH(1,1)
over the equity, others alternatives involve use the implied volatility of the previous prices of
options using a volatility surface model. Nevertheless, any alternative could have colinearity
issues with the previous regression that have to be address (maybe we will need to orthogonalize
some variables).

You might also like